Developmental Cell Article

Death Receptors DR6 and TROY Regulate Brain Vascular Development

Stephen J. Tam,1 David L. Richmond,1 Joshua S. Kaminker,2 Zora Modrusan,3 Baby Martin-McNulty,4 Tim C. Cao,4 Robby M. Weimer,4 Richard A.D. Carano,4 Nick van Bruggen,4 and Ryan J. Watts1,* 1Neurodegeneration Labs, Department of Neuroscience 2Bioinformatics 3Department of Molecular Biology 4Biomedical Imaging Genentech, 1 DNA Way, South San Francisco, CA 94080, USA *Correspondence: watts.ryan@gene.com DOI 10.1016/j.devcel.2011.11.018

SUMMARY (Tam and Watts, 2010). For example, how does Wnt signaling regulate CNS angiogenesis and barriergenesis? Which mole- Signaling events that regulate central nervous cules downstream of Wnt signaling drive BBB development? In system (CNS) angiogenesis and blood-brain barrier addition to the canonical Wnt signaling components, these (BBB) formation are only beginning to be elucidated. studies also identified tight junction proteins and transporters By evaluating the gene expression profile of mouse that may contribute to mature BBB function (Daneman et al., vasculature, we identified DR6/TNFRSF21 and 2009, 2010). TROY/TNFRSF19 as regulators of CNS-specific We reasoned that gene expression profiling the BBB vascula- ture at embryonic, neonatal, and adult stages would enable angiogenesis in both zebrafish and mice. Further- identification of signaling molecules driving CNS vascular devel- more, these two death receptors interact both genet- opment. Subsquently, we show that death receptors DR6 and ically and physically and are required for vascular TROY are enriched in CNS vasculature during embryogenesis, endothelial growth factor (VEGF)-mediated JNK acti- and accordingly drive angiogenesis and barriergenesis in zebra- vation and subsequent human brain endothelial fish and mouse model systems. Using an in vitro cell sprouting sprouting in vitro. Increasing beta-catenin levels in assay, we demonstrate that these two tumor necrosis factor brain endothelium upregulate DR6 and TROY, (TNF) receptor family members are required for endothelial indicating that these death receptors are down- sprouting events in a cell-autonomous fashion through VEGF- stream target genes of Wnt/beta-catenin signaling, mediated JNK activation of angiogenesis. Additionally, we find which has been shown to be required for BBB that DR6 and TROY genetically and physically interact, and development. These findings define a role for death may form a coreceptor complex at the BBB. Finally, we identify canonical Wnt/beta-catenin signaling as a key transcriptional receptors DR6 and TROY in CNS-specific vascular regulator of DR6 and TROY BBB expression, suggesting one development. possible mechanism by which Wnt/beta-catenin transcription factor activity drives BBB angiogenesis and barriergenesis.

INTRODUCTION RESULTS

The vasculature of the CNS is an example of uniquely differenti- Identification of Genes Enriched in Brain Vasculature ated organ-specific blood vessels. CNS blood vessels differen- To identify signaling molecules driving BBB development, matu- tiate to form the blood-brain barrier (BBB), which consists of ration, and maintenance, we evaluated the gene expression a complex network of intercellular tight and adherens junctions profile of mouse vasculature at three distinct developmental that form a molecular seal between adjacent endothelial cells time points (Figure 1A), corresponding to CNS angiogenesis to limit molecular exchange across CNS vessels. Accordingly, (E14.5), astrocytic endfeet contact with cortical endothelium the BBB has many transport mechanisms that enable specific (P7.5), and maintenance of a mature BBB (Adult). We utilized jettisoning of unwanted molecules out of the CNS while import- florescence-activated cell sorting (FACS) to isolate CD31-posi- ing molecules essential for brain function (Rubin and Staddon, tive, CD45-negative endothelial cells from mouse brain cortices, 1999; Tam and Watts, 2010). liver, and lung vasculature (Figure S1A available online), and The molecular cues that regulate BBB development have subsequently verified FACS-mediated CNS vascular purification remained elusive until recent findings described a role for the by qPCR analysis (Figure S1B). As expected, microarray analysis canonical Wnt/beta-catenin signaling cascade in CNS angio- identified several Wnt/beta-catenin target genes and other previ- genesis and barriergenesis (Daneman et al., 2009; Liebner ously characterized BBB-specific components to be enriched in et al., 2008; Stenman et al., 2008). These discoveries now BBB vasculature (Figure 1C; Figure S1)(Daneman et al., 2009; raise a number of important questions related to BBB biology Hallmann et al., 1995; Pardridge et al., 1990).

Developmental Cell 22, 403–417, February 14, 2012 ª2012 Elsevier Inc. 403 Developmental Cell Death Receptors Regulate CNS Vascular Development

A BRAIN VASCULAR DEVELOPMENT C BBB markers E14.5 P7.5 Adult Glut1 P-gp Meca32 Lef1 4 1 5 4 2 0 4 ** 2 ** ** -1 ** 0 ** 3 ** ** ** -2 0 -2 ** 2 ** -3

GLUT1 -4 1 -2 -4 ** EPAEPA EPAEPA EPAEPA EPAEPA

Log2 Expression Ratio Brain Liver/lung Brain Liver/lung Brain Liver/lung Brain Liver/lung

D Novel transcripts enriched at BBB DR6 TROY Spock2 Adcyap1r1 GFAP GFAP 6 8 6 2 ** 6 ** ** 4 ** ** ** 4 1 ** 4 ** ** ** 2 2 2 ** ** 0 0 0 0

P-gp EPAEPA EPAEPA EPAEPA EPAEPA Brain Liver/lung Brain Liver/lung Brain Liver/lung Brain Liver/lung Adora2b Dlx2 Tspn5 Gpr37L 1 4 4 5 B Z-Score 0 2 E (E14.5) 3 4

Log2 Expression Ratio 0 ** P (P7.5) -1 ** ** -2 ** 2 ** 3 A (Adult) -2 ** ** -2 0 2 4 ** -4 ** 1 2 -6 Brain Liver/lung -3 0 ** 1 E14.5 P7.5 Adult E14.5 P7.5 Adult EPAEPA EPAEPA EPAEPA EPAEPA Brain Liver/lung Brain Liver/lung Brain Liver/lung Brain Liver/lung

EPA E TNF receptor family members not enriched at BBB

EP Tnfrsf10b Tnfrsf11a Tnfrsf1a E (E14.5) 0.5 2 1.5 P (P7.5) PA 0.0 1.0 0 ** A (Adult) -0.5 0.5 EA -1.0 -2 ** ** 0.0 ** ** -1.5 -4 -0.5 -2.0 E EPAEPA EPAEPA EPAEPA

Log2 Expression Ratio Brain Liver/lung Brain Liver/lung Brain Liver/lung P A

Developmentally Regulated BBB-enriched Genes **p < 0.005, * p < 0.05

Figure 1. Blood-Brain Barrier Expression Profiling Identifies Genes Enriched in Brain Vasculature (A) Three blood-brain barrier (BBB) developmental time points identified by immunohistochemical analysis of GLUT1 expression and astrocyte localization in the mouse cortex. (GLUT1, endothelial cells; GFAP, astrocytes; P-gp, P-glycoprotein) Results representative of at least three independent experiments are shown. (B) Distinct developmentally regulated classes of BBB-specific transcripts were identified by microarray analysis. Samples are in columns; probes are in rows. Red and green color indicates over- or underexpression of probes in brain tissue relative to liver/lung tissue of the same stage, respectively. Developmental stage is indicated to the left of the heatmap (E, embryo; P, pup; A, adult). The probes shown in the heatmap are also presented in Table S1. (C) Expression profiling FACS-purified BBB and liver/lung vasculature shows enrichment of BBB markers Lef1, P-gp, and Glut1. (D) Selected transcripts enriched at the BBB. (E) TNF receptor family members Tnfrsf1a, 10b, and 11a are not preferentially expressed in CNS vasculature. Box and whisker plots indicate distribution of data and highlight median (horizontal line through box) and the 25th and 75th quartiles (box edges). Expression values were compared between brain and liver/lung tissue at each developmental stage. Statistical significance was defined as those stage-specific comparisons yielding an adjusted p value (**p < 0.005; *p < 0.05). Further details of statistical analysis are described in extended Experimental Procedures. See also Figure S1 and Table S1.

Microarray analysis across the BBB developmental timeline Further investigation of our gene expression data set yielded was able to resolve gene expression changes in a spatiotemporal several classes of BBB-specific transcripts: upregulated exclu- fashion (Figure 1B; Table S1; full microarray data set is publicly sively at E14.5, P7.5, adulthood, p7.5 and adulthood (e.g., available at the NCBI Gene Expression Omnibus). For instance, Tspn5 and Gpr37L), E14.5 and P7.5 (e.g., Adora2b and Dlx2), while the CNS-specific glucose transporter Glut1 is expressed in and across all three developmental time points (e.g., DR6, CNS at all three developmental time points, P-glycoprotein TROY, Spock2, and Adcyap1r1)(Figures 1B and 1D; Table S1). (P-gp) expression is highly upregulated at the BBB by postnatal Several of these transcripts have been previously described. day 7.5 (Figures 1A and 1C). Conversely, as Meca32 is only ex- For instance, both the proteoglycan SPOCK2 and homeobox pressed in fenestrated endothelium, transcript levels remained transcription factor DLX2 have been shown to be expressed in high in liver/lung vasculature at all time points and in embryonic CNS vessels and potentially drive CNS angiogenesis in a cell- brain vasculature, but are dramatically downregulated in brain autonomous fashion (Schnepp et al., 2005; Vasudevan et al., vasculature by P7.5 and adult time points (Figure 1C) (Hallmann 2008). Notably, the death receptors Tnfrsf21/DR6 and Tnfrsf19/ et al., 1995). TROY (Figure 1D; Figures S1D and S1E) were found to be highly

404 Developmental Cell 22, 403–417, February 14, 2012 ª2012 Elsevier Inc. Developmental Cell Death Receptors Regulate CNS Vascular Development

expressed in CNS-specific vasculature, while several TNF confocal live imaging of the CNS. Whereas control, Adcyap1r1, receptor family members including Tnfrsf10b, Tnfrsf11a, and Tspn5, and Tnfrsf1a morphant embryos excluded most of these Tnfrsf1a are not enriched in BBB vasculature (Figure 1E). tracers, as rhodamine-dextran was restricted to CtA vessel lumen while DAPI-stained nuclei were only found in CtA vessels, Zebrafish Candidate-Based Genetic Screen Identifies DR6, TROY, and Spock2 morphants exhibited dramatic leakage Regulators of CNS Angiogenesis and Barriergenesis of both tracers across the BBB, as evidenced by an increase in To understand the functional relevance of the BBB-specific tran- rhodamine-dextran signal and parenchymal nuclear staining by scripts identified in our expression profiling analysis, we em- DAPI (Figure 2A). Quantification of BBB leakage was accom- ployed a secondary screen utilizing the bony fish danio rerio plished by counting the number of DAPI-positive parenchymal (zebrafish) as a BBB developmental model system. Recently, ze- nuclei in the zebrafish brain. We observed a greater than 5-fold brafish was shown to have a functional BBB as early as 3 days increase in DAPI leakage across the BBB upon DR6, TROY postfertilization (dpf), expressing ZO-1 and CLAUDIN-5 tight and Spock2 knockdown compared to control embryos (Fig- junction protein in CNS endothelia (Jeong et al., 2008). To ure 2C). These findings support a role for DR6, TROY, and directly visualize CNS vasculature we utilized Kdrl-eGFP trans- SPOCK2 in CNS vascular development. genic zebrafish expressing green fluorescent protein in endothe- lial cells (Jin et al., 2005). We injected a solution containing DR6 Is Necessary for CNS Angiogenesis in Zebrafish 350 Da cationic DAPI and 10 kDa rhodamine-dextran into the We next set out to quantify zebrafish brain and trunk vasculature common cardinal vein of 3 dpf zebrafish embryos and found in response to DR6 loss of function. Consistent with the results dramatic retention of both dyes within CNS vasculature, from our screen, 2D quantification of 3 dpf zebrafish CNS and whereas peripheral vessels exhibited gross leakage into the trunk vasculature revealed that the number of hindbrain CtAs surrounding parenchymal space (Figures S2A and S2B) (Jeong per brain, but not trunk ISV density, was reduced upon DR6 et al., 2008). At this stage of development, blood vessels sprout- knockdown (Figures 3A and 3B). Likewise, fractional vessel ing from the primordial hindbrain channels (PHBC) and basilar heights were only reduced in the CNS. artery (BA) have formed an elaborate network of central arteries As vascular arborization is a product of coordinated sprouting, (CtA) in the brain (Figure 3C). Similar to the mammalian BBB, we branching, and lumenization (Adams and Alitalo, 2007), we found CNS-specific glucose transporter GLUT1 exclusively asked which of these processes are driven by DR6 by recon- expressed in this vascular network (Figure 1; Figure S3F) (Par- structing 3D confocal z stacks imaged from the developing dridge et al., 1990). zebrafish CNS vasculature and computationally measured the Importantly, the zebrafish genome contains orthologs for CtA density, branch density, and vessel diameter in response murine BBB-enriched molecules DR6, TROY, Spock2, to DR6-targeting morpholinos (Figure 3C). While both CtA vessel Adcyap1r1, and Tspn5 (Figures S5A and S5B; data not shown), diameter and branch density were not significantly altered upon and mRNA transcripts and protein corresponding to DR6 and DR6 knockdown, vascular density was decreased approxi- TROY expression were detectable (Figures S3A, S3E, S5C, mately 30% compared to control morpholino-injected embryos and S5D). Consistent with our murine expression profiling (Figures 3D). Furthermore, time-lapse imaging of 1.5–3 dpf data, zebrafish DR6 was detected in CNS vasculature (Figures embryos suggests that DR6 is required for initial CtA sprouting S3B and S3C). Therefore, we assessed CNS vasculature in 3 events from the BA and PHBC vascular networks, but not dpf embryos upon morpholino-mediated knockdown of these required to inhibit vascular regression or degeneration (Movies BBB-specific molecules at the developing zebrafish BBB S1, S2, S3, and S4). While DR6 knockdown phenotypes could (Figures S3A and S5C). Protein translation was sterically in- not be rescued upon DR6 overexpression, as death receptors hibited through translation-blocking antisense hybridization are acutely cytotoxic at high levels (Lavrik et al., 2007; Pan with mRNA start codon. Loss of function of BBB-enriched et al., 1998), the specificity of DR6 knockdown phenotypes DR6, TROY, Spock2, Adcyap1r1, and Tspn5 all resulted in robust was verified using another nonoverlapping translation blocking CNS-specific defects in vessel arborization, whereas trunk inter- morpholino, as well as a splice-blocking morpholino that leads segmental vessels (ISVs) were unperturbed (Figures 2A and 2B). to efficient pre-mRNA exon deletion in DR6 transcripts and Conversely, Tnfrsf1a (a TNF receptor family member that was not similar defects in CNS angiogenesis and barriergenesis (Fig- enriched at the BBB) knockdown embryos were indistinguish- ures S3D, S3E, S4E, and S4F). We conclude that DR6 is required able from control embryos. for CNS-specific angiogenesis, primarily through regulation of During vascular sprouting into the developing CNS many sprouting activity, but not through regression, anastamosis, or features specific to BBB function are initiated (Virgintino et al., lumenization. 2007). At the earliest stages of CNS angiogenesis, BBB compo- nents such as CLAUDIN-5 and GLUT1 are present along with DR6 Is Required for Proper CNS Vascular Density a physical barrier that can exclude small molecules from entering and Barrier Function in Mice the embryonic CNS (Daneman et al., 2009; Ek et al., 2006). We next examined DR6 expression in the developing mouse Therefore, our findings that DR6, TROY, SPOCK2, ADCYAP1R1, CNS vasculature. Immunohistochemistry (Figure 4A; Fig- and TSPN5 modulate CNS vascular development suggested ure S4 A), DR6.eGFP knockin expression (Figure S5F) and that these BBB-specific components may also regulate CNS in situ hybridization (Figure 4B) confirmed that DR6 is not only barriergenesis. To test this possibility, microangiographic injec- expressed in neurons, particularly commissural axons in the tions of leakage tracers DAPI and rhodamine-dextran were per- developing spinal cord, but also in CNS-specific vasculature formed on 3 dpf knockdown zebrafish embryos followed by (double labeling with GLUT1), suggesting that DR6 might be

Developmental Cell 22, 403–417, February 14, 2012 ª2012 Elsevier Inc. 405 iue2 erfihCniaeBsdGntcSre dnie euaoso N nignssadBarriergenesis and Angiogenesis CNS of Regulators Identifies Screen Genetic Candidate-Based Zebrafish 2. Figure opoioijcin AIpstv ri aecya uli(aecya AI :rgtpnl)wr iulyioae ycmuainlyrmvn brain removing computationally by isolated visually were panels) right A: 406 DAPI; (Parenchymal nuclei parenchymal brain in DAPI-positive transcripts BBB-enriched injection, of morpholino knockdown Morpholino-mediated B) and (A A eeomna Cell Developmental Tnfrsf1a_MO (-) Tspn5_MO Adcyap1r1_MO Spock2_MO TROY_MO DR6_MO Control_MO (-) 22 0–1,Fbur 4 2012 14, February 403–417, , DAPI Rhodamine-dextran Kdrl-eGFP Head PARENCHYMAL DAPI ª 02Esve Inc. Elsevier 2012 Tg(kdrl:egfp) et eetr euaeCSVsua Development Vascular CNS Regulate Receptors Death mro.A ecie in described As embryos. B DAPI Rhodamine-dextran Kdrl-eGFP Trunk xeietlProcedures Experimental *** p C eeomna Cell Developmental <0.005,** MO: MO: MO: DAPI positive nuclei DAPI positive nuclei MO: MO: DAPI positive nuclei DAPI positive nuclei MO: DAPI positive nuclei DAPI positive nuclei 10 20 30 40 10 20 30 40 10 20 30 40 10 20 30 40 10 20 30 40 10 20 30 40 0 0 0 0 0 0 oto Tnfrsf1a Control oto Spock2 Control oto Tspn5 Control oto Adcyap1r1 Control oto DR6 Control oto TROY Control BBB Leakage p <0.01,* olwn ng 4 a following , *** ** * * p <0.05 Developmental Cell Death Receptors Regulate CNS Vascular Development

necessary for murine CNS vascular patterning and/or barrier ation in neurons (Nikolaev et al., 2009), was identified in our function. expression profiling data to be modestly enriched in BBB vascu- In order to elucidate the functional relevance of DR6 enrich- lature exclusively during embryogenesis, and somewhat down- ment in mouse BBB vasculature, we characterized both vascular regulated in adult CNS vasculature, relative to liver/lung vascular density and barrier function in DR6 knockout mice. Ex vivo X-ray expression (Figure S4D). To determine whether APP drives DR6 microcomputed tomography (micro-CT) angiography was used activity during BBB development, we asked if APP and DR6 to quantitatively measure the whole-brain 3D cerebral vascular deletions share similar BBB developmental phenotypes. We structure. Importantly, we found that adult DR6 knockout mice found that while APP loss of function in zebrafish and mice ex- exhibit 30% decrease in brain vascular density compared to hibited subtle vascular abnormalities (Figures S4E, S4G, and wild-type littermates (Figures 4C and 4D; Figure S4B). This S4H), APP loss of function did not result in BBB leakage (Figures defect is similar to vascular density reduction in zebrafish S4E–S4G). We conclude that APP is not required for CNS bar- embryos upon DR6 knockdown (Figure 3D; Figure S3D). riergenesis and therefore likely does not drive DR6 activity during Our findings that DR6 is required for proper CNS vascular BBB development. density in the mature mouse brain, and drives both CNS angio- genesis and barriergenesis during zebrafish development sug- Cell-Autonomous Regulation of Vascular Sprouting gest that DR6 may also regulate CNS barrier function in the adult by DR6 mouse. To test this possibility, we evaluated whether DR6 In order to examine the contribution of endothelial cell-derived knockout mice also have a disrupted BBB by quantifying CNS DR6 toward BBB development in more detail, we utilized an extravasation of intravenously injected Evans blue dye, as previ- in vitro human brain microvascular endothelial cell (HBMEC) ously described (Eliceiri et al., 1999). Even though DR6 KO adult sprouting assay that we established whereby primary HBMECs mice are viable, these animals exhibited an approximately 2-fold were coated onto cytodex beads and allowed to sprout in a 3D increase in Evans blue dye leakage compared to wild-type litter- fibrin matrix. In a similar fashion to DR6, Neuropilin-1 (NRP1) mate controls (Figure 4E), which is similar in magnitude to has been demonstrated to guide the development of both the barrier deficits observed after intracortical delivery of VEGF (Fig- nervous and vascular systems (Gu et al., 2003; Tam and Watts, ure S4C), a well-known cytokine that induces vascular perme- 2010). Accordingly, function-blocking antibodies targeting NRP1 ability. Therefore, we conclude that DR6 contributes to CNS strongly inhibited HBMEC vascular remodeling and sprouting, vascular morphogenesis and barrier function in mice. which confirmed HBMEC sprouting activity to be NRP1-depen- To determine if DR6 loss-of-function phenotypes in adult mice dent (Figure 5A) (Pan et al., 2007). Importantly, two distinct func- may be a consequence of dysregulated CNS angiogenesis and tion-blocking antibodies targeting DR6 directly inhibited the barriergenesis during development, we investigated possible number of sprouting HBMECs and their mean sprout length to embryonic phenotypes. Indeed, DR6 knockout embryos ex- a similar extent as anti-NRP1 antibody treatment (Figure 6A; Fig- hibited both forebrain-specific hemorrhaging events (Figure 4F) ure S6C). Furthermore, siRNA-mediated knockdown of DR6 as well as an 15% increase in sulfo-NHS-biotin leakage across yielding >70% reduction in transcript and protein levels (Figures the BBB upon transcardiac perfusion (Figure 4G). Furthermore, S6A and S6B), also significantly reduced HBMEC sprout length embryonic hindbrain radial vessel density was significantly and cell number (Figure 5B), albeit to a lesser extent than anti- reduced in DR6 knockout embryos (Figures 4H and 4I). To deter- DR6 function-blocking antibodies. These results suggest that mine if these phenotypes arise as a consequence of DR6 specif- during BBB development, DR6 directs brain vascular develop- ically in endothelial cells, we generated a floxed exon2 DR6 allele ment and barriergenesis through regulation of sprouting in a and crossed it to a Tie2-Cre mouse driver line to remove DR6 cell-autonomous fashion. from endothelium in vivo. Similar to our observations in DR6 knockout embryos (Figures 4F, 4H, and 4I), we find that endo- TROY Is Necessary for CNS Angiogenesis thelial-specific DR6 deletion in DR6flox/flox;Tie2-Cre embryos Based on our DR6 results and data from our zebrafish screen, we recapitulated the DR6 null phenotype with forebrain-specific speculated that another death receptor, TROY, may work in hemorrhaging events at E11.5 and reduced hindbrain radial concert with DR6 to regulate CNS vascular development. vessel density (Figures 4J–4L). Furthermore, we observed that Consistent with our zebrafish screening results (Figure 2), 2D both the DR6 knockout and the endothelial-specific DR6 condi- quantification of 3 dpf zebrafish CNS and trunk vasculature re- tional knockout show a reduction in tight junction component vealed that the number of hindbrain CtAs per brain, but not trunk ZO-1 protein levels (Figure S7B). These early developmental ISV density, was reduced upon TROY knockdown (Figures 5C BBB phenotypes are consistent with the idea that DR6 is an and 5D). Likewise, fractional vessel heights were only reduced essential regulator of BBB development in a cell-autonomous in the CNS. While brain vascular density decreased by 30% fashion. in response to TROY-targeting morpholino, there were no signif- Intriguingly, amyloid precursor protein (APP), a candidate DR6 icant defects with respect to CtA branch density or vessel diam- ligand that was proposed to drive developmental axon degener- eter (Figure 5E). Time-lapse imaging of 1.5–3 dpf embryos also vasculature-derived DAPI signal from the DAPI channel. Results representative of at least three independent experiments are shown. Endothelial cells are green, DAPI is blue, rhodamine-dextran is red. MO, morpholino. (C) Quantification of DAPI leakage across the zebrafish BBB. n = 6 (control_MO), n > 6 (DR6_MO, TROY_MO, Spock2_MO, Adcyap1r1_MO, Tspn5_MO, and Tnfrsf1a_MO). Data are presented as the mean ± standard error of the mean (SEM) (***p < 0.005; **p < 0.01; *p < 0.05). See also Figure S2.

Developmental Cell 22, 403–417, February 14, 2012 ª2012 Elsevier Inc. 407 Developmental Cell Death Receptors Regulate CNS Vascular Development

Control MO DR6 MO A

Height Head Number

Height Trunk

Number

B Brain Vasculature Trunk Vasculature

12 0.5 14 1.2 12 10 * 0.4 *** 1 8 10 0.8 0.3 8 6 0.6 0.2 6 4 4 0.4 2 0.1 0.2

Number of CtAs 2

0 0 ISV Density (mm-1) 0 0 Fractional Height CtAs MO: Control DR6 Control DR6 Control DR6 Fractional Height ISVs Control DR6

C BA PCS

PHBC

D 3-Dimensional Brain Vasculature

1.2 12 4 600 *** 3 0.8 8 400 2 0.4 4 200 1

0 0 0 0 Vascular Density ( µ m/nl) Vascular Branch Density (100 µ m)-1 MO: Control DR6 Control DR6 Diameter ( µ m) CtA Average Control DR6 Control DR6 Stdev of CtA Diameter ( µ m) Stdev of CtA E DLV MCeV MtA R2 BCA PHBC

R3 R1 BA PHBC ***p < 0.005, * p < 0.05

Figure 3. Zebrafish Blood-Brain Barrier Requires DR6 for CNS Angiogenesis (A) Translation-blocking morpholino-mediated DR6 knockdown leads to CNS-specific angiogenic defects in Tg(kdrl:egfp) embryos. Each embryo was treated with 4 ng morpholino. Results representative of at least three independent experiments are shown. MO, morpholino; CtA, central artery; ISV, intersegmental vessels. (B) 2D quantification of 3 dpf zebrafish CNS and trunk vasculature. n = 36 (control_MO), n = 15 (DR6_MO), see Experimental Procedures, Zebrafish Micro- angiographic Imaging and Quantification, for calculation of the fractional height of CtAs and ISVs.

408 Developmental Cell 22, 403–417, February 14, 2012 ª2012 Elsevier Inc. Developmental Cell Death Receptors Regulate CNS Vascular Development

suggested that TROY is required for initial CtA sprouting events CNS vessel formation and barrier function (Figures 6C and 6D). but not for degenerative developmental processes during CNS Embryos coinjected with both morpholinos at these concentra- vessel formation (Movies S5 and S6). We verified morpholino- tions displayed synergistic CNS-specific vascular defects and mediated knockdown using a splice-blocking morpholino, which barrier leakage. In other words, simultaneous suboptimal DR6 inhibits splicing events in TROY transcripts and resulted in and TROY knockdown unmasked otherwise silent BBB-specific similar defects in CNS angiogenesis and barriergenesis phenotypes that resulted in an enhanced genetic interaction (Figure S5D). similar to yeast gene function mapping studies identifying gene To further establish that TROY plays a role in CNS vascular deletion pairs that uncover sick/lethal synthetic phenotypes biology we evaluated both TROY knockout mice phenotypes (Tong et al., 2001). The identification of both BBB-specific and TROY regulation of angiogeneic sprouting in the HBMEC genetic and physical interactions between DR6 and TROY bead assay. Adult TROY knockout mice exhibited a modest suggests that binding partners DR6 and TROY function in unison increase in Evans blue dye BBB leakage compared to wild- to contribute to CNS angiogenesis and barriergenesis during type littermate controls (Figure 5F). Furthermore, siRNA-medi- development. ated knockdown of TROY yielding >85% reduction in transcript levels (Figure S6A) significantly reduced HBMEC sprout length DR6 and TROY Are Required for VEGF-Mediated JNK and cell number (Figure 5B). Therefore, similar to DR6, TROY is Activation and Subsequent Human Brain Endothelial required for CNS-specific angiogenesis, primarily through regu- Sprouting lation of sprouting activity, but not through regression, anasta- Mediated largely through proangiogenic factor VEGF binding to mosis, or lumenization. cognate receptor VEGFR2, sprouting angiogenesis requires the orchestration of a number of cellular events that are integrated DR6 and TROY Physically Interact by molecular signals including cellular migration via p38 (Rous- As DR6 and TROY loss of function shared similar CNS vascula- seau et al., 1997), proliferation via ERK (Rousseau et al., 1997; ture developmental defects in zebrafish and sprouting defects Takahashi et al., 1999), and survival/vascular permeability via in vitro, we hypothesized that DR6 and TROY may form func- PI3-kinase/AKT (Chen et al., 2005; Gerber et al., 1998; Six tional coreceptors. Specific TNF receptors have been shown et al., 2002). Interestingly, wild-type or dominant-negative to heteromultimerize with other family members, as well as other ERK2 overexpression triggers enhancement or suppression of signaling receptor families. For instance, TNF receptor family JNK activation, respectively, which demonstrates that VEGF- member p75 inhibits axon regeneration through Nogo-66 mediated ERK phosphorylation activates downstream JNK (Pe- receptor (NgR1) binding (Wang et al., 2002). TNF receptor dram et al., 1998). Accordingly, JNK activation has since been TROY can also form a receptor complex with NgR1 and function- identified as a positive regulator of human umbilical vein endothe- ally replace p75 in neurons (Park et al., 2005; Shao et al., 2005). lial cell (HUVEC) angiogenesis (Uchida et al., 2008; Wu et al., Indeed, when transfected into COS7 cells, TROY can be coim- 2006). Indeed, we found that VEGF strongly stimulates JNK acti- munoprecipitated with NgR1 (Figure 6A). Notably, TROY ap- vation in HBMECs, as evidenced by increased phospho-JNK peared to homo-oligomerize upon transfection as HA-tagged levels (Figure 6E). In order to verify the hypothesis that JNK acti- TROY could pulldown flag-tagged TROY. Consistent with our vation is required for efficient brain vessel sprouting, two struc- prediction, transfected DR6 strongly coimmunoprecipitated turally distinct and highly selective JNK inhibitors (JNK-V and with TROY to an even greater degree than compared to positive JNK-VIII) were individually tested in the HBMEC bead-sprouting control NgR1 (Figure 6A). Furthermore, both anti-DR6 function- assay (Gaillard et al., 2005; Szczepankiewicz et al., 2006). Upon blocking antibodies that were used to inhibit HBMEC sprouting addition of each JNK inhibitor, we found robust inhibition of were able to partially block DR6/TROY binding (Figure 6B). both HBMEC sprout length and sprouting cell number to a similar Considering that purified DR6 extracellular domains did not degree as DR6 and TROY knockdown (Figure 6F). directly bind to TROY-expressing 293 cells (Figure S6D), we While death receptors are typically thought to functionally propose a model whereby DR6 and TROY physically interact interact with the cellular apoptosis machinery, different cellular through their cytoplasmic or transmembrane domains to form environments can shift death receptors toward mediating func- a functional receptor complex in CNS vascular development. tions distinct from, or even against cell death (Ashkenazi and Dixit, 1998). For instance, TNF binding to death receptor DR6 and TROY Genetically Interact TNFR1 activates NF-kB signaling which in turn suppresses Having shown that DR6 and TROY physically interact, we next TNF-induced apoptosis (Beg and Baltimore, 1996). Only upon examined their potential genetic interactions. We coinjected sufficient suppression of NF-kB signaling can TNF-TNFR1 DR6 and TROY targeting morpholinos at suboptimal concentra- induce cell death. Additionally, death receptor TNFRSF6/CD95 tions that individually caused no detectable defects in zebrafish has recently been shown to promote tumor cell proliferation

(C) 3D reconstructions of zebrafish hindbrain vasculature were used to create a vascular mask (second panel from the left) to determine vascular density and average CtA diameter. Filament tracing of the vessel-derived eGFP signal (first and second panels from the right) enabled branch density calculations. (D) 3D quantification of 3 dpf zebrafish hindbrain vasculature. Standard deviation of CtA diameter is a metric of irregularity in vessel lumenization. n=4 (control_MO), n = 6 (DR6_MO).

(E) Zebrafish brain volume calculation diagram. The volume of the half brain was approximated as V=R1R2R3. MtA, metencephalic artery; MCeV, middle cerebral vein; DLV, dorsal longitudinal vein; BCA, basal communicating artery; PHBC, primordial hindbrain channel; BA, basilar artery. Data are presented as the mean ± SEM (***p < 0.005; *p < 0.05). See also Figure S3 and Movies S1, S2, S3, and S4.

Developmental Cell 22, 403–417, February 14, 2012 ª2012 Elsevier Inc. 409 Developmental Cell Death Receptors Regulate CNS Vascular Development

A C DR6.WT E11.5 DR6.KO E11.5 DR6.WT DR6.KO GLUT1 DR6 GLUT1 DR6 DR6 DR6

D E Adult Brain Vascularity Adult BBB Leakage

3 3 *

2 * 2

1 1 % vascular/brain volume B 0 DR6.WT E14.5 0 DR6.WT DR6.KO Relative BBB leakage (Evans blue) DR6.WT DR6.KO GLUT1 GLUT1 DR6 J DR6loxp/loxp; DR6 Control Tie2-Cre E11.5 F G DR6.WT DR6.KO E18.5 BBB Leakage

1.2 **

DR6loxp/loxp; K Control Tie2-Cre 0.8 GLUT1 GLUT1 E11.5 0.4 E14.5

0 Relative BBB leakage (NHS-biotin) DR6.WT DR6.KO

H I E14.5 Brain Vascularity L E14.5 Brain Vascularity DR6.WT DR6.KO 160 250 -2) -2)

GLUT1 GLUT1 * 200 120 *** 150 80 100 E14.5 40 50 Hindbrain Vessel Density (mm Hindbrain Vessel 0 Density (mm Hindbrain Vessel 0 DR6.WT DR6.KO Control DR6loxp/loxp; ***p < 0.005 ; **p < 0.01; *p < 0.05 Tie2-Cre

Figure 4. DR6 Is Required for Proper CNS Vascular Density and Barrier Function in Mice in a Cell-Autonomous Fashion (A) DR6 protein expression in CNS commissural axons and GLUT1-positive vasculature. Immunohistochemical staining for DR6 protein exhibited both neuronal and vascular signal in spinal cord (top panels) and forebrain (bottom panels) tissue from E11.5 embryos. This signal is specific to DR6 protein, as no antibody

410 Developmental Cell 22, 403–417, February 14, 2012 ª2012 Elsevier Inc. Developmental Cell Death Receptors Regulate CNS Vascular Development

through downstream JNK activation (Chen et al., 2010). The dual and tight junction protein localization within CNS vasculature functions of CD95 can be explained by the observation that (Figure 4A; Figures S3F, S5E, and S7), suggesting that the CD95-mediated apoptosis requires 1,000-fold higher protein Wnt/beta-catenin pathway is likely not downstream of DR6 levels compared to its nonapoptotic activity (Lavrik et al., and TROY signaling. However, we found that DR6 and TROY 2007). Therefore, the cellular consequences of death receptor knockouts exhibit reduced ZO-1 protein levels, which suggests activation of downstream signaling pathways function in a that these death receptors may regulate brain vascular develop- context-dependent manner. The death receptors DR6 and ment partially through tight-junction component modulation TROY were initially identified as JNK-activating members of (Figure S7B). the TNFR superfamily that promote cell death (Eby et al., 2000; We next investigated whether the Wnt signaling pathway Pan et al., 1998). Interestingly, JNK2/3 double knockouts have drives CNS angiogenesis and barrier formation in part by modu- been previously described to exhibit gross vascular defects lating DR6 and TROY expression at the BBB. To examine the (Laukeviciene et al., 2006). Therefore, we reasoned that DR6 possibility that these two death receptors are downstream effec- and TROY might promote CNS angiogenesis through nonapop- tors that partially account for Wnt/beta-catenin involvement in totic JNK activation. BBB development, we measured DR6 and TROY transcript In order to delineate potential signaling crosstalk between levels in response to recombinant hWnt3a ligand, hWnt3a condi- VEGF-VEGFR2 and these BBB-specific death receptors, proan- tioned media, and GSK3beta inhibitor SB216763 in HBMECs. In giogenic signaling pathways were examined in HBMECs trans- a similar fashion to canonical Wnt target gene Axin2, DR6 and fected with siRNA against either DR6 or TROY followed by TROY were significantly upregulated in response to two different VEGF addition. VEGF-mediated activation of ERK, AKT, or p38 forms of exogenous Wnt3a ligand, as well as pharmacological signaling was not modified upon DR6 or TROY knockdown (Fig- stabilization of beta-catenin by SB216763 (Figures 7A, 7B, and ure 6E). However, consistent with previous cell culture studies 7D). Furthermore, Wnt3a-mediated transcriptional upregulation demonstrating JNK activation upon DR6 or TROY overexpres- of these two BBB-specific death receptors was completely sion (Eby et al., 2000; Pan et al., 1998), we found that VEGF- reversible upon siRNA-mediated beta-catenin knockdown or mediated JNK phosphorylation was reduced upon abrogation addition of the pan-Wnt ligand binding soluble Fzd8-Fc protein of DR6 or TROY expression in human brain endothelial cells (Fig- (Figures 7C and 7D) (DeAlmeida et al., 2007; Gong et al., 2010). ure 6E, right). These results are consistent with the finding that The Wnt/beta-catenin developmental pathway employs DR6 and TROY are required for VEGF-mediated JNK activation expression feedback loops to enable temporal autoregulation and subsequent human brain endothelial sprouting. through Wnt-responsive gene expression (Logan and Nusse, 2004). In light of our findings that DR6 and TROY are Wnt- DR6 and TROY Are Downstream of Canonical responsive genes, we reasoned that DR6 and TROY may also Wnt/Beta-Catenin Signaling in CNS Vasculature participate in Wnt signaling autoregulation. A prediction of this Recently, canonical Wnt/beta-catenin signaling was identified as idea is that death receptor loss of function would modulate brain a key regulator essential for CNS-specific angiogenesis, barrier- vascular Wnt target gene expression levels. Indeed, expression genesis, and GLUT1 expression at the BBB in a cell-autonomous analysis of FACS-purified brain vasculature revealed that DR6 fashion (Daneman et al., 2009; Liebner et al., 2008; Stenman knockout BBB vasculature exhibit marked reduction of two et al., 2008). Intriguingly, we found that DR6 and TROY knock- canonical Wnt target gene transcripts, Apcdd1 and Axin2 (Fig- outs maintain high levels of GLUT1 expression, low levels of ure 7E). However, given that Wnt-driven GLUT1 expression is vesicular-protein PLVAP, normal PDGFRbeta-positive pericyte largely unaffected upon DR6 or TROY loss of function (Figure 4A;

staining was apparent in DR6.KO embryos (right panels). Results representative of at least three independent experiments are shown. WT, wild-type; KO, knockout. (B) DR6 mRNA expression in Glut1-positive vasculature. Dual-target two-color fluorescent in situ hybridization was used to detect DR6 (red) and Glut1 (green) transcripts in mouse E14.5 brains. Results representative of at least three independent experiments are shown. (C) MicroCT angiographic surface renderings of adult mouse brain vasculature. 2D projections of DR6.WT and DR6.KO brains are shown. Results representative of six independent experiments are shown. (D) Quantification of microCT CNS vascular network. Vascular density was calculated as a function of vascular volume normalized by brain volume. n = 6 (DR6.WT), n = 6 (DR6.KO). (E) Evans blue leakage assay detects barrier defects in adult mice. n = 7 (DR6.WT), n = 8 (DR6.KO). (F) E11.5 DR6.KO embryos exhibit forebrain-specific hemorrhaging (observed in 60% embryos tested; arrow). Results representative of at least three inde- pendent experiments are shown. (G) Fixable biotin transcardiac perfusion assay detects barrier defects in E18.5 mouse embryos. n = 12 (DR6.WT), n = 12 (DR6.KO). (H) DR6.KO embryos exhibit reduced hindbrain vascular density at E14.5. Confocal images of flat-mounted hindbrains representative of at least nine independent experiments are shown. (I) Quantification of E14.5 mouse hindbrain vascular density. Vascular density was calculated as a function of number of GLUT1-positive hindbrain radial vessels normalized by field area. n = 12 (DR6.WT), n = 11 (DR6.KO). (J) E11.5 DR6flox/flox;Tie2-Cre embryos exhibit forebrain-specific hemorrhaging (observed in 80% embryos tested; arrow). Results representative of at least three independent experiments are shown. (K) E11.5 DR6flox/flox;Tie2-Cre embryos exhibit reduced hindbrain vascular density at E14.5. Confocal images of flat-mounted hindbrains representative of at least nine independent experiments are shown. (L) Quantification of E14.5 mouse hindbrain vascular density. Vascular density was calculated as described above. n = 15 (control), n = 11 (DR6flox/flox;Tie2-Cre). Data are presented as the mean ± SEM (***p < 0.005; **p < 0.01; *p < 0.05). See also Figure S4.

Developmental Cell 22, 403–417, February 14, 2012 ª2012 Elsevier Inc. 411 Developmental Cell Death Receptors Regulate CNS Vascular Development

A Human Brain Endothelial Sprouting B Human Brain Endothelial Sprouting 200 Control Anti-DR6 Actin 80 80 ** 20 DAPI *** ** 100 ** *** *** *** 40 40 10 ** ** Sprouting Cell Count Sprouting Cell Count ** Relative Sprout Length Relative Sprout Length

0 0 0 0 Control Anti-DR6A Anti-DR6B Anti-NRP1 Control Anti-DR6A Anti-DR6B Anti-NRP1 siControl siTROY siDR6 siControl siTROY siDR6

CD F Brain Vasculature Trunk Vasculature Control MO TROY MO 12 0.5 14 1.2 Adult BBB Leakage 10 0.4 *** 12 1 * 8 *** 10 0.8 0.3 8 6 0.6 6 0.2 2 4 4 0.4 0.1 Head 2 2 0.2 Number of CtAs Fractional Height ISVs ISV Density (mm-1) 0 Fractional Height CtAs 0 0 0 * MO:Control TROY Control TROY Control TROY Control TROY 1 E 3-Dimensional Brain Vasculature

0 TROY.WT TROY.KO 1.2 12 4

600 Relative BBB leakage (Evans blue) ** 3 0.8 8 Trunk 400 2 0.4 4 200 1

Vascular Density ( µ m/nl) Vascular 0 0 0 0 Average CtA Diameter ( µ m) CtA Average Branch Density (100 µ m)-1

MO:Control TROY Control TROY Control TROY Diameter ( µ m) Stdev of CtA Control TROY

***p < 0.005; **p < 0.01; * p < 0.05

Figure 5. DR6 and TROY Are Required for Proper CNS Vascular Density and Barrier Function through Brain Endothelial Angiogenic Sprouting in a Cell-Autonomous Fashion (A) Anti-DR6 function-blocking antibodies reduce HBMEC sprouting. Actin is green, nuclei are blue (left). Results representative of at least three independent experiments are shown. Quantification of sprout length (middle) and number of sprouting cells (right) are shown. (B) siRNA-mediated knockdown of DR6 and TROY inhibits HBMEC sprouting. (C) Translation-blocking morpholino-mediated TROY knockdown leads to CNS-specific angiogenic defects. Each Tg(kdrl:egfp) zebrafish embryo was treated with 4 ng morpholino. Results representative of at least three independent experiments are shown. (D) 2D quantification of 3 dpf zebrafish vasculature. n = 36 (control_MO), n = 25 (TROY_MO). (E) 3D quantification of 3 dpf zebrafish hindbrain vasculature. n = 5 (control_MO), n = 6 (TROY_MO). (F) Evans blue leakage assay detects barrier defects in adult mice. n = 16 (TROY.WT), n = 18 (TROY.KO). Data are presented as the mean ± SEM (***p < 0.005; **p < 0.01; *p < 0.05). See also Figure S5 and Movies S5 and S6.

Figures S3F and S5E), DR6-mediated autoregulation of Wnt signaling toward either proliferation or caspase-mediated cell signaling at the BBB may only have modest functional relevance death through context-dependent molecular cues. Alternatively, and therefore requires more detailed investigation. In conclu- the cellular functions of these receptors may be predetermined sion, we find that DR6 and TROY expression is regulated by within specific cell types. For instance, DR6 and TROY might canonical Wnt/beta-catenin signaling at the BBB, thereby driving not be able to engage the apoptotic machinery in endothelial BBB-specific activity of these death receptors to contribute cells. toward CNS angiogenesis and barriergenesis during develop- Although Wnt/beta-catenin signaling is indispensible for ment (Figure 7F). BBB-specific GLUT1 expression, we find that DR6 and TROY are not. How then do DR6 and TROY contribute to CNS DISCUSSION vessel and barrier formation at the molecular level? Are the barrier defects associated with DR6 and TROY loss of function A Progrowth/Morphogenic Role for Death Receptors a direct consequence of vascular developmental defects, or The death receptor CD95 has been shown to function in direct do these death receptors drive CNS angiogenesis and barrier- opposition to, or even entirely independent of necrotic cell death genesis independently? Although DR6, TROY, and Spock2 (Chen et al., 2010). Similarly, we find that DR6 and TROY knockdown resulted in both vascular and barrier defects, promote BBB development, independent of their canonical pro- BBB-enriched molecules Adcyap1r1 and Tspn5 knockdown degenerative or apoptotic activities. We therefore propose that exhibited gross brain vascular malformations in the absence nonapoptotic activity mediated by death receptors may be of barrier leakage phenotypes, suggesting that impaired CNS a general property inherent to these signaling molecules. Specif- vascular development does not inevitably inhibit BBB ically, death receptors may act as cell fate switches that amplify formation.

412 Developmental Cell 22, 403–417, February 14, 2012 ª2012 Elsevier Inc. Developmental Cell Death Receptors Regulate CNS Vascular Development

ACB mDR6-flag D TROY-HA + + + + mTROY-HA Control_MO DR6_MO low

TROY-flag + A B Kdrl-eGFP BBB Leakage Rhodamine-dextran DR6-flag + DAPI + 80 * NgR1-flag ControlAnti-DR6Anti-DR6 IP: flag IP: flag IB: HA IB: HA 60 IP: flag IP: flag IB: flag IB: flag TROY_MO low DR6_MO + TROY_MO 40 input IB: input IB: HA HA DAPI positive nuclei input IB: 20 flag input IB: flag 1 * ** 0.5 0 MO: Control DR6 TROY DR6 low + low low TROY low TROY co-IP’d TROY 0 E F VEGF VEGF VEGF VEGF Human Brain Endothelial Sprouting

120 40 Neg control siControl siTROY siDR6 Neg control Neg control siControl siTROY siDR6 Neg control siControl siTROY siDR6 siControl siTROY siDR6

80 P-ERK P-AKT P-p38 P-JNK ** *

20 *** 40 ERK AKT p38 JNK *** Sprouting Cell Count Relative Sprout Length 8 2 0 0 DMSO JNK V JNK VIII 1 DMSO JNK V JNK VIII 2 4 1

0 0 0 0 Relative P-AKT/AKT Relative P-ERK/ERK Relative P-p38/p38 Relative P-JNK/JNK ***p<0.005; **p < 0.01; *p < 0.05

Figure 6. DR6 and TROY Physically and Genetically Interact and Are Required for VEGF-Mediated JNK Activation (A) DR6 physically interacts with TROY. Epitope-tagged receptors were overexpressed in COS-7 cells and flag-tagged TROY bait was immunoprecipitated. Input was 1% total cellular lysate used in pulldown. IP, immunoprecipitation; IB, immunoblot. Results representative of at least three independent experiments are shown. (B) Anti-DR6 function-blocking antibodies, which inhibit human brain endothelial sprouting, also reduce DR6 and TROY physical interaction. Pulldown assays were carried out as described above with the addition of anti-DR6 antibodies four hours after transfection. Relative TROY coimmunoprecipitation was calculated by normalization to both TROY-HA input and DR6-flag pulldown levels. Results representative of at least three independent experiments are shown. (C) Suboptimal concentrations of DR6- and TROY-targeting morpholinos unmask genetic interactions. Each embryo was treated with 4 ng of each splice- blocking morpholino. Endothelial cells are green, DAPI is blue, rhodamine-dextran is red. Results representative of at least three independent experiments are shown. (D) Quantification of DAPI leakage reveals DR6 and TROY genetic interaction. n = 8 (control_MO), n = 8 (DR6_MO), n = 8 (TROY_MO), n = 8 (DR6_MO and TROY_MO). (E) VEGF activates JNK in HBMECs and requires DR6 and TROY expression. Relative protein phosphorylation was calculated by normalization to total protein levels. Results representative of at least three independent experiments are shown. (F) JNK activity is required for HBMEC sprouting. 1 mM JNK inhibitor V or VIII was used. Data are presented as the mean ± SEM (***p < 0.005; **p < 0.01; *p < 0.05). See also Figure S6.

Death Receptor Crosstalk with VEGF-VEGFR2 Signaling cell apoptosis and inhibit angiogenesis through modulating In addition to the signaling kinases p38, ERK, and AKT, we find VEGF signaling by inhibition of AKT (Caporali et al., 2008). that the VEGF-VEGFR2 pathway activates JNK in human brain endothelial cells in a DR6 and TROY-dependent fashion. We Death Receptor Regulation in CNS Vasculature postulate that death receptors directly interface with the down- It remains to be determined if DR6 and TROY function in a ligand- stream signaling elements of the VEGF-VEGFR pathway in order independent fashion, or if DR6 and TROY act as coreceptors for to effectively modulate sprouting angiogenesis. This signaling an unidentified ligand to mediate CNS vascular biology. Notably, axis would then allow for the integration of signaling input from TNF receptor family members including TNFRSF1A/TNFR, both VEGF ligands and death receptors. Consistent with this TNFRSF10B/DR5, and TNFRSF5/CD40 can assemble into pre- hypothesis, p75 activity has been shown to promote endothelial formed trimer complexes in the absence of cognate ligand

Developmental Cell 22, 403–417, February 14, 2012 ª2012 Elsevier Inc. 413 Developmental Cell Death Receptors Regulate CNS Vascular Development

A BCsiControl Control siControl + Wnt3a Control siBeta-catenin + Wnt3a Wnt3a GSK3beta inhibitor 3 *** 1.6 *** * *** * 3 1.2 2 *** 2 *** ** 0.8 # ** # 1 1 0.4 Relative Expression (A.U.) Relative Expression (A.U.) Relative Expression (A.U.) # 0 0 0 TROY DR6 Axin2 TROY DR6 Axin2 TROY DR6 Axin2 Beta-catenin

D E F WNT Control Wnt3a-CM DR6.WT Brain Wnt3a-CM + Fzd8-Fc DR6.KO Brain DR6.WT Liver/Lung BETA-CATENIN DR6.KO Liver/Lung *** 4 1 *** * *** DR6/TROY ???

VEGF VEGFR2 JNK 2 *** # #

Relative Expression (A.U.) ANGIOGENESIS BARRIERGENESIS

0 Relative Expression (A.U.) 0 TROY DR6 Axin2 Beta-catenin DR6 Axin2 Apcdd1

***p < 0.005; **p < 0.01; * p< 0.05; #p < 0.005 +/- sibeta-catenin or Fzd8-Fc

Figure 7. DR6 and TROY Are Downstream of Canonical Wnt/Beta-Catenin Signaling in CNS Vasculature (A) Wnt3a ligand stimulates DR6 and TROY expression in human brain endothelial cells. Recombinant human Wnt3a (200 ng/ml) was added to HBMECs and RNA was analyzed 24 hr postligand addition by qPCR. (B) Increased beta-catenin levels stimulate DR6 and TROY expression in human brain endothelial cells. GSK3beta inhibitor SB216763 (20 mM) was added to HBMECs and analyzed as described above. (C) Recombinant Wnt3a-mediated DR6 and TROY expression is dependent on beta-catenin. HBMEC RNA was analyzed as above 24 hr after Wnt3a addition ± siRNA targeting beta-catenin. (D) Wnt3a-conditioned media stimulates DR6 and TROY expression in a ligand-dependent fashion. HBMEC RNA was analyzed as above 24 hr after Wnt3a-CM addition ± pan-Wnt binding Fzd8-Fc. (E) Wnt signaling in embryonic mouse brain vasculature is attenuated upon DR6 knockout. CD31+CD45- vasculature was FACS-purified as in Figure S1A and analyzed as described above. (F) Model for death receptors DR6 and TROY function in BBB development. Wnt ligands secreted by the embryonic neuroepithelium stabilize beta-catenin protein levels in CNS endothelial cells which subsequently upregulates, among many other target genes, DR6 and TROY expression. DR6, and possibly TROY, participate in a Wnt/beta-catenin positive feedback loop whereby signaling is amplified by these death receptors through an unknown mechanism. Through JNK activation, these receptors engage in crosstalk with the VEGF/VEGFR2 signaling pathway to drive CNS angiogenesis and possibly barriergenesis. Data are presented as the mean ± SEM (***p < 0.005; **p < 0.01; *p < 0.05; #p < 0.005 comparing Wnt3a ± Fzd8-Fc or sibeta-catenin). See also Figure S7. binding (Chan et al., 2000), while TNFRSF8/CD30 overexpres- et al., 2010; Klapholz-Brown et al., 2007; Qiu et al., 2010; sion has been shown to drive downstream NF-kB signaling in a Shin et al., 2010). Accordingly, we find that DR6 and TROY are ligand-independent fashion (Horie et al., 2002). Thus, it is downstream target genes of canonical Wnt/beta-catenin possible that DR6 and TROY function at the BBB as a ligand- signaling in human brain endothelium, which is exclusively acti- independent signaling unit. vated in CNS-specific endothelial cells during development. Importantly, we found that DR6 and TROY are coregulated at Considering that loss of Wnt/beta-catenin signaling results in the BBB. In addition to exhibiting similar loss-of-function pheno- more severe BBB developmental defects when compared to types, these two death receptors physically and genetically DR6 and TROY loss of function, it must be the case that DR6 interact, and have overlapping expression patterns. Recently, and TROY work in concert with other Wnt target genes to fully both in silico transcription factor binding prediction and expres- regulate CNS angiogenesis and barriergenesis. Nevertheless, sion profiling studies to identify Wnt/beta-catenin target genes in our results provide a mechanistic explanation of how DR6 human fibroblasts, dermal papilli, and stromal stem cells found and TROY coexpression is regulated at the BBB. Specifically, both DR6 and TROY as potential downstream effectors (Ho¨ dar DR6 and TROY are expressed in CNS endothelial cells through

414 Developmental Cell 22, 403–417, February 14, 2012 ª2012 Elsevier Inc. Developmental Cell Death Receptors Regulate CNS Vascular Development

neuroepithelium-derived Wnt ligands and subsequent induction tors, including JNK inhibitor V (Calbiochem AS601245), JNK VIII inhibitor of Wnt/beta-catenin transcriptional target genes. It is through (Calbiochem 420135), and GSK3beta inhibitor SB216763 (Tocris 1616), were the expression of these particular death receptors, as well as added 24 hr before cell lysis. The bead outgrowth assay was adapted from (Nakatsu et al., 2003). In brief, hydrated collagen-coated dextran beads (Amer- other unidentified components, that Wnt/beta-catenin signaling sham Cytodex-3 17-0485-01) were coated with HBMECs (300 cells/bead) by regulates organ-specific formation and differentiation of CNS shaking cells every 10–30 min for 3 hr and allowed to grow in a flask overnight vasculature. in 10 ml fresh media. Fibrinogen (25 mg/ml; Sigma F8630) was made in EGM2 and diluted 10-fold in bead medium. Bead medium consists of EGM2, 1:5 EXPERIMENTAL PROCEDURES human skin fibroblast-conditioned media (ATCC CCL-110 in EGM2), 2 ng/ml hVEGF, and 2 ng/ml HGF. Two hundred HBMEC-coated beads/ml released Zebrafish Microangiographic Imaging and Quantification from the flask bottom by aggressive tapping were washed and added into Microangiography and live imaging of brain vascularization were adapted from fibrinogen/bead medium mixture and 1 ml pipetted into wells of 12 well plate  (Jeong et al., 2008; Lawson and Weinstein, 2002). For microangiographic with 10 ml thrombin (Sigma T9549). After 20 min in 37 C incubator, 2 ml assessment of BBB integrity, morpholino-injected embryos were anesthetized bead media was slowly added to the fibrin gel matrix along with function- at 3.25 dpf in 0.04% tricaine pH 7 (Sigma E10521), and injected with 12 nl of blocking antibodies or small molecule inhibitors. Cells were fixed 1% PFA in  12.5 mg/ml rhodamine-dextran and 0.85 mg/ml DAPI mix into the common PBS overnight at 4 C, washed twice in PBS, and stained in PBS with cardinal vein. Embryos were mounted in 1% low-melt agarose in Danieau 1:10,000 Hoescht and 1:250 Phalloidin-488 conjugate (Invitrogen A12379) overnight at 4C. After two PBS washes, confocal z stacks of the entire plate buffer (58 mM NaCl, 0.7 mM KCl, 0.4 mM MgSO4, 0.6 mM Ca[NO3]2, and 5 mM HEPES [pH 7.6]) containing tricaine in a glass-bottom 96-well plate, was acquired using the ImageXpress Micro (Molecular Devices). Mean sprout and z stacks (100 mm thick, 3 mm spacing between sagittal planes, three lengths were calculated by manual tracing and the number of sprouting cells line averages) were collected within 30 min of injection. All images were were counted automatically using ImageJ. acquired using a laser scanning confocal microscope (TCS SP5, Leica) with 3 a 500 mW Argon-Ion laser (LASOS) and a 20 air objective (Apochromat 0.7 Animals NA, Leica). Mice, zebrafish, and their embryos were handled in accordance with Genen- The integrity of the BBB and extent of brain vascularization were quantified tech IACUC guidelines. in control and morpholino-injected fish from two dimensional (2D) maximum z projections and three dimensional (3D) renderings of the z stacks. 2D maximum z projections were created using ImageJ (developed by the National SUPPLEMENTAL INFORMATION Institutes of Health), and the number of DAPI positive brain parenchymal nuclei was manually counted for each condition (n > 6 animals per condition). DAPI- Supplemental Information includes seven figures, Supplemental Experimental positive brain parenchymal nuclei were visually isolated using Adobe Photo- Procedures, one table, and six movies and can be found with this article online shop CS5 subtraction blending whereby DAPI+GFP+ double-positive signal at doi:10.1016/j.devcel.2011.11.018. (DAPI labeling of brain vasculature) was computationally removed from the DAPI channel. 2D z projections were also used to quantify the fractional height ACKNOWLEDGMENTS of the central arteries (CtA) penetrating into the embryo brain by normalizing CtA height against the height of the dorsal longitudinal vein (DLV) above the We thank members of the Watts laboratory, Anatoly Nikolaev, Sidney Hsieh, junction of the posterior communicating segment (PCS) and the basilar artery Joe Lewcock, Kyle Niessen, Bob Liu, David Simon, Jeremy Burton, Philip (BA). Similarly, the fractional height of ISVs was calculated relative to the sepa- Vitorino, James Ernst, and John Brady Ridgway for kindly providing reagents, ration of the dorsal aorta from the dorsal longitudinal anastomotic vessel. advice, and stimulating discussions. We thank Murat Yaylaoglu for in situ The brain vasculature was reconstructed in 3D by first deconvolving the z hybridization experiments, Rupak Neupane, Laurie L. Gilmour, Chungkee stacks using AutoQuant X (AutoQuant Imaging) and then 3D rendering with Poon, and James E. Cupp for FACS sorting and analysis, Jed Ross for Imaris (Bitplane). Quantification of total length and volume of CtAs in the left ex vivo mouse brain imaging techniques, Cecile Chalouni and Laszlo Komuves hindbrain, and number of branches within this network of CtAs were performed for confocal microscopy guidance, and Janice Maloney for subcloning TROY using the Surface Tool and Filament Tracking packages in Imaris. Both surface expression constructs. All authors are paid employees of Genentech. rendering and filament tracking were conducted using default parameters whenever possible; however, manual correction of the filament tracking Received: June 8, 2011 results, using the raw fluorescence signal as a guide, was often necessary. Revised: October 7, 2011 Vessel and branch density were subsequently calculated by normalizing Accepted: November 14, 2011 against the volume of the left hindbrain. Brain volume was approximated as Published online: February 13, 2012 V=R1R2R3, where R1 is the length of the BA, R2 is the height of the DLV above the junction of the PCS and BA, and R3 is the distance from to the primordial hindbrain channel (PHBC) to the junction of the PCS and BA, measured REFERENCES perpendicular to the BA (Figure 3E). Live imaging of vascularization in the zebrafish brain was carried out as Adams, R.H., and Alitalo, K. (2007). Molecular regulation of angiogenesis and above with a few modifications: 5–20 embryos were dechorionated at 1–3 lymphangiogenesis. Nat. Rev. Mol. Cell Biol. 8, 464–478. dpf and embedded in 1% low-melt agarose in Danieau buffer containing tri- Ashkenazi, A., and Dixit, V.M. (1998). Death receptors: signaling and modula-  caine and 0.03 mg/ml N-phenylthiourea (PTU; Sigma P7629) at 28.5 C. The tion. Science 281, 1305–1308. coordinates of each embryo were stored using a motorized stage (Ma¨ rzha¨ user Beg, A.A., and Baltimore, D. (1996). An essential role for NF-kappaB in pre- Wetzlar) and time-lapse images were acquired every 10–30 min for up to 12 hr. venting TNF-alpha-induced cell death. Science 274, 782–784. The embryo stage and frame rate and total duration of imaging are provided with the supplemental movies. Caporali, A., Pani, E., Horrevoets, A.J., Kraenkel, N., Oikawa, A., Sala-Newby, G.B., Meloni, M., Cristofaro, B., Graiani, G., Leroyer, A.S., et al. (2008). Primary Human Brain Microvascular Endothelial Cell Culture and Neurotrophin p75 receptor (p75NTR) promotes endothelial cell apoptosis Bead Outgrowth Assay and inhibits angiogenesis: implications for diabetes-induced impaired neovas- HBMECs from Cell Systems (ACBRI 376) were maintained in Lonza EGM2 Sin- cularization in ischemic limb muscles. Circ. Res. 103, e15–e26. gleQuots (CC-3162). Dharmacon ON-TARGETplus SMARTpool siRNAs were Chan, F.K., Chun, H.J., Zheng, L., Siegel, R.M., Bui, K.L., and Lenardo, M.J. transfected in primary cells using Dharmafect transfection reagent #1 (2000). A domain in TNF receptors that mediates ligand-independent receptor (T-2001) according to the manufacturer’s instructions. Small molecule inhibi- assembly and signaling. Science 288, 2351–2354.

Developmental Cell 22, 403–417, February 14, 2012 ª2012 Elsevier Inc. 415 Developmental Cell Death Receptors Regulate CNS Vascular Development

Chen, J., Somanath, P.R., Razorenova, O., Chen, W.S., Hay, N., Bornstein, P., Laukeviciene, A., Brecht, S., Kevelaitis, E., and Herdegen, T. (2006). Enhanced and Byzova, T.V. (2005). Akt1 regulates pathological angiogenesis, vascular contractility of small blood vessels in JNK knockout mice. Eur. J. Pharm. Sci. maturation and permeability in vivo. Nat. Med. 11, 1188–1196. 29, 335–339. Chen, L., Park, S.M., Tumanov, A.V., Hau, A., Sawada, K., Feig, C., Turner, Lavrik, I.N., Golks, A., Riess, D., Bentele, M., Eils, R., and Krammer, P.H. J.R., Fu, Y.X., Romero, I.L., Lengyel, E., and Peter, M.E. (2010). CD95 (2007). Analysis of CD95 threshold signaling: triggering of CD95 (FAS/APO- promotes tumour growth. Nature 465, 492–496. 1) at low concentrations primarily results in survival signaling. J. Biol. Chem. Daneman, R., Agalliu, D., Zhou, L., Kuhnert, F., Kuo, C.J., and Barres, B.A. 282, 13664–13671. (2009). Wnt/beta-catenin signaling is required for CNS, but not non-CNS, Lawson, N.D., and Weinstein, B.M. (2002). In vivo imaging of embryonic angiogenesis. Proc. Natl. Acad. Sci. USA 106, 641–646. vascular development using transgenic zebrafish. Dev. Biol. 248, 307–318. Daneman, R., Zhou, L., Agalliu, D., Cahoy, J.D., Kaushal, A., and Barres, B.A. Liebner, S., Corada, M., Bangsow, T., Babbage, J., Taddei, A., Czupalla, C.J., (2010). The mouse blood-brain barrier transcriptome: a new resource for Reis, M., Felici, A., Wolburg, H., Fruttiger, M., et al. (2008). Wnt/beta-catenin understanding the development and function of brain endothelial cells. PLoS signaling controls development of the blood-brain barrier. J. Cell Biol. 183, ONE 5, e13741. 409–417. DeAlmeida, V.I., Miao, L., Ernst, J.A., Koeppen, H., Polakis, P., and Rubinfeld, Logan, C.Y., and Nusse, R. (2004). The Wnt signaling pathway in development B. (2007). The soluble wnt receptor Frizzled8CRD-hFc inhibits the growth of and disease. Annu. Rev. Cell Dev. Biol. 20, 781–810. teratocarcinomas in vivo. Cancer Res. 67, 5371–5379. Nakatsu, M.N., Sainson, R.C., Aoto, J.N., Taylor, K.L., Aitkenhead, M., Pe´ rez- Eby, M.T., Jasmin, A., Kumar, A., Sharma, K., and Chaudhary, P.M. (2000). del-Pulgar, S., Carpenter, P.M., and Hughes, C.C. (2003). Angiogenic sprout- TAJ, a novel member of the tumor necrosis factor receptor family, activates ing and capillary lumen formation modeled by human umbilical vein endothelial the c-Jun N-terminal kinase pathway and mediates caspase-independent cells (HUVEC) in fibrin gels: the role of fibroblasts and Angiopoietin-1. cell death. J. Biol. Chem. 275, 15336–15342. Microvasc. Res. 66, 102–112. Ek, C.J., Dziegielewska, K.M., Stolp, H., and Saunders, N.R. (2006). Functional Nikolaev, A., McLaughlin, T., O’Leary, D.D., and Tessier-Lavigne, M. (2009). effectiveness of the blood-brain barrier to small water-soluble molecules in APP binds DR6 to trigger axon pruning and neuron death via distinct caspases. developing and adult opossum (Monodelphis domestica). J. Comp. Neurol. Nature 457, 981–989. 496, 13–26. Pan, G., Bauer, J.H., Haridas, V., Wang, S., Liu, D., Yu, G., Vincenz, C., Eliceiri, B.P., Paul, R., Schwartzberg, P.L., Hood, J.D., Leng, J., and Cheresh, Aggarwal, B.B., Ni, J., and Dixit, V.M. (1998). Identification and functional char- D.A. (1999). Selective requirement for Src kinases during VEGF-induced acterization of DR6, a novel death domain-containing TNF receptor. FEBS angiogenesis and vascular permeability. Mol. Cell 4, 915–924. Lett. 431, 351–356. Gaillard, P., Jeanclaude-Etter, I., Ardissone, V., Arkinstall, S., Cambet, Y., Pan, Q., Chanthery, Y., Liang, W.C., Stawicki, S., Mak, J., Rathore, N., Tong, Camps, M., Chabert, C., Church, D., Cirillo, R., Gretener, D., et al. (2005). R.K., Kowalski, J., Yee, S.F., Pacheco, G., et al. (2007). Blocking neuropilin- Design and synthesis of the first generation of novel potent, selective, and 1 function has an additive effect with anti-VEGF to inhibit tumor growth. in vivo active (benzothiazol-2-yl)acetonitrile inhibitors of the c-Jun N-terminal Cancer Cell 11, 53–67. kinase. J. Med. Chem. 48, 4596–4607. Pardridge, W.M., Boado, R.J., and Farrell, C.R. (1990). Brain-type glucose Gerber, H.P., Dixit, V., and Ferrara, N. (1998). Vascular endothelial growth transporter (GLUT-1) is selectively localized to the blood-brain barrier. factor induces expression of the antiapoptotic proteins Bcl-2 and A1 in Studies with quantitative western blotting and in situ hybridization. J. Biol. vascular endothelial cells. J. Biol. Chem. 273, 13313–13316. Chem. 265, 18035–18040. Gong, Y., Bourhis, E., Chiu, C., Stawicki, S., DeAlmeida, V.I., Liu, B.Y., Park, J.B., Yiu, G., Kaneko, S., Wang, J., Chang, J., He, X.L., Garcia, K.C., and Phamluong, K., Cao, T.C., Carano, R.A., Ernst, J.A., et al. (2010). Wnt iso- He, Z. (2005). A TNF receptor family member, TROY, is a coreceptor with Nogo form-specific interactions with coreceptor specify inhibition or potentiation receptor in mediating the inhibitory activity of myelin inhibitors. Neuron 45, of signaling by LRP6 antibodies. PLoS ONE 5, e12682. 345–351. Gu, C., Rodriguez, E.R., Reimert, D.V., Shu, T., Fritzsch, B., Richards, L.J., Kolodkin, A.L., and Ginty, D.D. (2003). Neuropilin-1 conveys semaphorin and Pedram, A., Razandi, M., and Levin, E.R. (1998). Extracellular signal-regulated VEGF signaling during neural and cardiovascular development. Dev. Cell 5, protein kinase/Jun kinase cross-talk underlies vascular endothelial cell growth 45–57. factor-induced endothelial cell proliferation. J. Biol. Chem. 273, 26722–26728. Hallmann, R., Mayer, D.N., Berg, E.L., Broermann, R., and Butcher, E.C. Qiu, W., Hu, Y., Andersen, T.E., Jafari, A., Li, N., Chen, W., and Kassem, M. (1995). Novel mouse endothelial cell surface marker is suppressed during (2010). Tumor necrosis factor receptor superfamily member 19 (TNFRSF19) differentiation of the blood brain barrier. Dev. Dyn. 202, 325–332. regulates differentiation fate of human mesenchymal (stromal) stem cells through canonical Wnt signaling and C/EBP. J. Biol. Chem. 285, 14438–14449. Ho¨ dar, C., Assar, R., Colombres, M., Aravena, A., Pavez, L., Gonza´ lez, M., Martı´nez, S., Inestrosa, N.C., and Maass, A. (2010). Genome-wide identifica- Rousseau, S., Houle, F., Landry, J., and Huot, J. (1997). p38 MAP kinase acti- tion of new Wnt/beta-catenin target genes in the human genome using vation by vascular endothelial growth factor mediates actin reorganization and CART method. BMC Genomics 11, 348. cell migration in human endothelial cells. Oncogene 15, 2169–2177. Horie, R., Watanabe, T., Morishita, Y., Ito, K., Ishida, T., Kanegae, Y., Saito, I., Rubin, L.L., and Staddon, J.M. (1999). The cell biology of the blood-brain Higashihara, M., Mori, S., Kadin, M.E., and Watanabe, T. (2002). Ligand-inde- barrier. Annu. Rev. Neurosci. 22, 11–28. pendent signaling by overexpressed CD30 drives NF-kappaB activation in Schnepp, A., Komp Lindgren, P., Hu¨ lsmann, H., Kro¨ ger, S., Paulsson, M., and Hodgkin-Reed-Sternberg cells. Oncogene 21, 2493–2503. Hartmann, U. (2005). Mouse testican-2. Expression, glycosylation, and effects Jeong, J.Y., Kwon, H.B., Ahn, J.C., Kang, D., Kwon, S.H., Park, J.A., and Kim, on neurite outgrowth. J. Biol. Chem. 280, 11274–11280. K.W. (2008). Functional and developmental analysis of the blood-brain barrier Shao, Z., Browning, J.L., Lee, X., Scott, M.L., Shulga-Morskaya, S., Allaire, N., in zebrafish. Brain Res. Bull. 75, 619–628. Thill, G., Levesque, M., Sah, D., McCoy, J.M., et al. (2005). TAJ/TROY, an Jin, S.W., Beis, D., Mitchell, T., Chen, J.N., and Stainier, D.Y. (2005). Cellular orphan TNF receptor family member, binds Nogo-66 receptor 1 and regulates and molecular analyses of vascular tube and lumen formation in zebrafish. axonal regeneration. Neuron 45, 353–359. Development 132, 5199–5209. Shin, H., Kwack, M.H., Shin, S.H., Oh, J.W., Kang, B.M., Kim, A.A., Kim, J., Klapholz-Brown, Z., Walmsley, G.G., Nusse, Y.M., Nusse, R., and Brown, P.O. Kim, M.K., Kim, J.C., and Sung, Y.K. (2010). Identification of transcriptional (2007). Transcriptional program induced by Wnt protein in human fibroblasts targets of Wnt/beta-catenin signaling in dermal papilla cells of human scalp suggests mechanisms for cell cooperativity in defining tissue microenviron- hair follicles: EP2 is a novel transcriptional target of Wnt3a. J. Dermatol. Sci. ments. PLoS ONE 2, e945. 58, 91–96.

416 Developmental Cell 22, 403–417, February 14, 2012 ª2012 Elsevier Inc. Developmental Cell Death Receptors Regulate CNS Vascular Development

Six, I., Kureishi, Y., Luo, Z., and Walsh, K. (2002). Akt signaling mediates VEGF/ Systematic genetic analysis with ordered arrays of yeast deletion mutants. VPF vascular permeability in vivo. FEBS Lett. 532, 67–69. Science 294, 2364–2368. Stenman, J.M., Rajagopal, J., Carroll, T.J., Ishibashi, M., McMahon, J., and Uchida, C., Gee, E., Ispanovic, E., and Haas, T.L. (2008). JNK as a positive McMahon, A.P. (2008). Canonical Wnt signaling regulates organ-specific regulator of angiogenic potential in endothelial cells. Cell Biol. Int. 32, 769–776. assembly and differentiation of CNS vasculature. Science 322, 1247–1250. Vasudevan, A., Long, J.E., Crandall, J.E., Rubenstein, J.L., and Bhide, P.G. Szczepankiewicz, B.G., Kosogof, C., Nelson, L.T., Liu, G., Liu, B., Zhao, H., (2008). Compartment-specific transcription factors orchestrate angiogenesis Serby, M.D., Xin, Z., Liu, M., Gum, R.J., et al. (2006). Aminopyridine-based gradients in the embryonic brain. Nat. Neurosci. 11, 429–439. c-Jun N-terminal kinase inhibitors with cellular activity and minimal cross- Virgintino, D., Girolamo, F., Errede, M., Capobianco, C., Robertson, D., kinase activity. J. Med. Chem. 49, 3563–3580. Stallcup, W.B., Perris, R., and Roncali, L. (2007). An intimate interplay between Takahashi, T., Ueno, H., and Shibuya, M. (1999). VEGF activates protein kinase precocious, migrating pericytes and endothelial cells governs human fetal C-dependent, but Ras-independent Raf-MEK-MAP kinase pathway for DNA brain angiogenesis. Angiogenesis 10, 35–45. synthesis in primary endothelial cells. Oncogene 18, 2221–2230. Wang, K.C., Kim, J.A., Sivasankaran, R., Segal, R., and He, Z. (2002). P75 Tam, S.J., and Watts, R.J. (2010). Connecting vascular and nervous system interacts with the Nogo receptor as a co-receptor for Nogo, MAG and development: angiogenesis and the blood-brain barrier. Annu. Rev. OMgp. Nature 420, 74–78. Neurosci. 33, 379–408. Wu, G., Luo, J., Rana, J.S., Laham, R., Sellke, F.W., and Li, J. (2006). Tong, A.H., Evangelista, M., Parsons, A.B., Xu, H., Bader, G.D., Page´ , N., Involvement of COX-2 in VEGF-induced angiogenesis via P38 and JNK path- Robinson, M., Raghibizadeh, S., Hogue, C.W., Bussey, H., et al. (2001). ways in vascular endothelial cells. Cardiovasc. Res. 69, 512–519.

Developmental Cell 22, 403–417, February 14, 2012 ª2012 Elsevier Inc. 417